On fixing a bug (and all its siblings) with a forward looking view

We run into a strange situation deep in the guts of RavenDB. A cluster command (the backbone of of RavenDB is coordinating action in a distributed cluster) failed because of an allocation failure. That is something that we are ready for, since RavenDB is a robust system which handles such memory allocation failures.